Conversations should be vectorized

crossid / accessbot

Streamline resource access grants with AI-Powered chatbot

https://www.crossid.io

Other

2 stars 0 forks source link

Open asaf opened 6 months ago

asaf commented 6 months ago

ATM if the conversation was ended with an actual grant then vectorize it

There is some level of confidence about the quality of the request if the conversation has approved via a data owner.
ThumbUp/ThumbDown is not very useful because it takes time for the user to check if grants are positive, we should consider in the future a delayed thumbup/down (e.g., 1 day after granting)
Let infosec vote the conversations.

ErezSha commented 5 months ago

To sum up our conversation from 26.6:

should be vectorized: 1.1 data owner conversations that required the recommender and were approved. 1.2 data owner conversations that reached him without a recommendation in the first place.
should not be vectorized: 2.1 any denied recommendation 2.2 recommendations that were approved without debate by the data owner

I'll have to figure out how to get the agent to make a request to the app's data owner without a recommendation (for the second point)